Hal R. Varian

Recommender systems

Recommender systems assist and augment this natural social process. In a typical recommender system people provide recommendations as inputs, which the system then aggregates and directs to appropriate recipients. In some cases the primary transformation is in the aggregation; in others the system’s value lies in its ability to make good matches between the recommenders and those seeking recommendations. The developers of the first recommender system, Tapestry [1], coined the phrase “collaborative filtering” and several others have adopted it. We prefer the more general term “recommender system” for two reasons. First, recommenders may not explictly collaborate with recipients, who may be unknown to each other. Second, recommendations may suggest particularly interesting items, in addition to indicating those that should be filtered out. This special section includes descriptions of five recommender systems. A sixth article analyzes incentives for provision of recommendations. Figure 1 places the systems in a technical design space defined by five dimensions. First, the contents of an evaluation can be anything from a single bit (recommended or not) to unstructured textual annotations. Second, recommendations may be entered explicitly, but several systems gather implicit evaluations: GroupLens monitors users’ reading times; PHOAKS mines Usenet articles for mentions of URLs; and Siteseer mines personal bookmark lists. Third, recommendations may be anonymous, tagged with the source’s identity, or tagged with a pseudonym. The fourth dimension, and one of the richest areas for exploration, is how to aggregate evaluations. GroupLens, PHOAKS, and Siteseer employ variants on weighted voting. Fab takes that one step further to combine evaluations with content analysis. ReferralWeb combines suggested links between people to form longer referral chains. Finally, the (perhaps aggregated) evaluations may be used in several ways: negative recommendations may be filtered out, the items may be sorted according to numeric evaluations, or evaluations may accompany items in a display. Figures 2 and 3 identify dimensions of the domain space: The kinds of items being recommended and the people among whom evaluations are shared. Consider, first, the domain of items. The sheer volume is an important variable: Detailed textual reviews of restaurants or movies may be practical, but applying the same approach to thousands of daily Netnews messages would not. Ephemeral media such as netnews (most news servers throw away articles after one or two weeks) place a premium on gathering and distributing evaluations quickly, while evaluations for 19th century books can be gathered at a more leisurely pace. The last dimension describes the cost structure of choices people make about the items. Is it very costly to miss IT IS OFTEN NECESSARY TO MAKE CHOICES WITHOUT SUFFICIENT personal experience of the alternatives. In everyday life, we rely on

On the private provision of public goods

Abstract We consider a general model of the non-cooperative provision of a public good. Under very weak assumptions there will always exist a unique Nash equilibrium in our model. A small redistribution of wealth among the contributing consumers will not change the equilibrium amount of the public good. However, larger redistributions of wealth will change the set of contributors and thereby change the equilibrium provision of the public good. We are able to characterize the properties and the comparative statics of the equilibrium in a quite complete way and to analyze the extent to which government provision of a public good ‘crowds out’ private contributions.

Predicting the Present with Google Trends

In this paper we show how to use search engine data to forecast near-term values of economic indicators. Examples include automobile sales, unemployment claims, travel destination planning and consumer confidence.

Predicting the Present with Google Trends: PREDICTING THE PRESENT WITH GOOGLE TRENDS

In this paper we show how to use search engine data to forecast near�?term values of economic indicators. Examples include automobile sales, unemployment claims, travel destination planning and consumer confidence.

Redistributive taxation as social insurance

Abstract The modern literature on nonlinear optimal taxation treats differences in income as being due to unobserved differences in ability. A striking result of this assumption is that high income agents should face a zero marginal tax rate. In this paper I assume that differences in observed income are due to exogenous differences in luck. Hence the optimal redistributive tax involves trading off the benefits due to ‘social insurance’ with the costs due to reduced incentives. I derive the optimal forms for linear and nonlinear taxes, and compute some algebraic and numeric examples. Typically high income individuals will face quite high marginal tax rates.

Non-parametric Tests of Consumer Behaviour

This paper shows how to test demand data for consistency with maximization, homotheticity, various forms of separability, and a rationing model without making any assumptions concerning the parametric form of underlying demand or utility functions.

System Reliability and Free Riding

System reliability often depends on the effort of many individuals, making reliability a public good. It is well-known that purely voluntary provision of public goods may result in a free rider problem: individuals may tend to shirk, resulting in an inefficient level of the public good. How much effort each individual exerts will depend on his own benefits and costs, the efforts exerted by the other individuals, and the technology that relates individual effort to outcomes. In the context of system reliability, we can distinguish three prototype cases.

BUYING, SHARING AND RENTING INFORMATION GOODS*

Information goods such as books, journals, computer software, music and videos can be copied, shared, resold, or rented. When such opportunities for sharing are present, the content producer will generally sell a smaller amount at a higher price which may increase or decrease profits. I identify three circumstances where profits increase: (1) when the transactions cost of sharing is less than the marginal cost of production; (2) when content is viewed only a few times and transactions costs of sharing are low; and (3) when a sharing market provides a way to segment high-value and low-value users. Copyright 2000 by Blackwell Publishing Ltd

Sequential contributions to public goods

Abstract I examine games involving private contributions to a public good and show that less of the public good will be supplied if agents move sequentially than if they move simultaneously. If the agents bid for the right to move first, the agent who values the public good least will win. If each agent chooses the rate at which he will subsidize the other agents contributions, the subsidies that support the Lindahl allocation are the unique equilibrium outcome. I also describe two related subsidy-setting games that yield Lindahl allocations in n -person games with general utility functions.

Non-parametric analysis of optimizing behavior with measurement error

We consider how one might test observed choice data for consistency with optimizing models in the presence of measurement error. We derive an appropriate test statistic and conduct case study involving cost minimization behavior by electric utility plants.